Abstract
The debate on the link between ethnicity and violence has been raging in political science literature since the end of the Cold War. Often, cross-country quantitative studies dismissed the importance of ethnic heterogeneity as a source of violent conflict. How the patterns of ethnic settlement within a country affect the severity of violence, though, has not yet been studied through similar techniques. In this essay, we build and analyse a data set of major violence-related variables collected at the local level during the 1992–1995 war in Bosnia-Herzegovina. What emerges is that the local distribution of the population, in terms of the number and relative size of the groups, is a key factor in explaining the intensity of violence in the Bosnian municipalities.
Acknowledgements
The authors would like to thank Vanja Gavran, Franjo Topić, and two anonymous referees for their comments. The article is the product of a joint work. For the sole purpose of formal attribution, sections 1, 2, 4, and 5 are written by Stefano Costalli; sections 3, 6 and 7 are written by Francesco N. Moro.
Notes
1. A recent summary of the debate on ethnicity can be found in Jenkins (Citation2008); with more direct reference to ethnic conflict and violence see Laitin (Citation2007) and Brubaker and Laitin (Citation1998).
2. As is clear from the study of Buhaug, Cederman and Rød (2008), the two explanations are in fact intertwined.
3. The year 1995 is included in its entirety in our data set.
4. The dimension of ethnic identities used to be so important that only 5.5 per cent of the inhabitants of Bosnia-Herzegovina considered themselves as ‘Yugoslavs’.
5. The relationship between these two different measures of ethnic composition can also be captured thanks to descriptive statistics: low levels of fractionalization and polarization refer to the same areas so that the municipalities with the ten lowest levels of fractionalization and polarization are exactly the same. On the contrary, the indexes strongly diverge when we look at their highest values, causing different predictions concerning the highest levels of violence.
6. Melander (Citation2009) and Slack and Doyon (Citation2001) represent similar efforts to study the local dimension of violence in Bosnia-Herzegovina, but they use very different research designs and no multivariate analysis.
8. The fractionalization index can be expressed with the formula: FRAG = 1−Σsij 2, where s stands for the percentage of the ethnic group i in the municipality j. The index of polarization is instead expressed by the formula: POLAR=1−Σ(0.5 – πi)2 πi/0.25, where πi represents the percentage of an ethnic group in a given municipality.
9. The index of correlation is 0.859.
10. In this case, the level of violence is measured by the number of deaths/population of the municipality.
11. The case of Srebrenica has been included in the other models because it does not alter the results in relevant ways.
12. The variable that accounts for the level of income per capita in 1991 has mean = 4890.899 and Std. Dev.=1006.351. The values are expressed in Dinars.
Additional information
Notes on contributors
Stefano Costalli
STEFANO COSTALLI is Post-Doctoral Fellow at the Catholic University, Milan
Francesco N. Moro
FRANCESCO N. MORO is Post-Doctoral Fellow at the Italian Institute of Human Sciences.